Download Big Island Discussions II 08 03 06 - Alt

Big Island Discussions II 08 03 06
Telescope Structures and Active Primary Mirror Support Systems
I returned to Hilo for discussions with the Gemini and Subaru engineers on
March 3rd and 4th. The purpose of this informal memo is to summarize those
discussions.
At Gemini Headquarters, from the left, Chris Carter (Gemini Controls), Daigo Tomono (Subaru
Mechanical), Mike Sheehan (Gemini Mechanical), and Russ Genet (Cal Poly Univ.).
Mike Sheehan (Gemini Mechanical Engineering) and I discussed telescope
structures. Mike drew a force versus frequency diagram for wind disturbances
on telescopes with the more or less steady forces below 0.2 Hz, and the
turbulent forces higher than 0.2 Hz where resonances can create problems for
control system stability. In general, if one makes the telescope stiff enough to
avoid resonances in the higher frequency regime, the structure will be more than
robust enough for the constant wind forces. Thus in structural design, one has to
be primarily concerned with higher frequency resonances.
We also discussed tip tilting and rotational forces on the soil through the pier for
systems not on bedrock. These movements are not sensed by the telescope’s
position encoders, of course, but show up as motions that must be removed by
active star-tracking tip/tilt systems. Soil springiness may be a particularly difficult
problem for small portable science/astrophotography telescopes. We discussed
whether or not placing a number of very heavy weights on the bottom, nonmoving portion of a portable telescope might improve this situation. Such
weights might actually lower the overall telescope/soil resonant frequency, but
make it less likely to be excited.
1
We then discussed dampening. The very stiff telescope structures required to
avoid low frequency resonances generally end up having very little dampening,
i.e. are high “Q.” Mike demonstrated the vibration of two strips of aluminum.
One vibrated for quite some time. The other, which had the same amount of
aluminum in it, but was in two half thicknesses separated by a thin absorbing
material, did not vibrate at all. The samples were supplied by CSA Engineering
in Mt. View, California. Ty Safreno has mentioned this firm before. They
specialize in solving (and avoiding) vibration issues.
To inject some fun into our discussions, I asked Mike for some design ideas for a
2-meter equivalent telescope with four 1 meter, thin spherical mirrors. Our alt-az
development group occasionally considers various options for extending small
telescope apertures beyond 1 meter while still maintaining low costs and light
weights. Such telescopes might be especially useful for time series photometry
in the optical and near IR, and for fiber fed optical spectroscopy.
For the telescope we considered, the four spherical mirrors would, individually,
be f/6, and in an aggregate f/3. The mirrors would be coaligned, and cofocused,
but would not be cophased, so their resolution could not exceed that of a single
mirror (not much of an issue at typical lowland sites).
Dave Rowe has done some very preliminary optical design work that suggests
that a 2 or 3 element spherical lens corrector could provide acceptable images
(for typical lowland sites) over fields, for instance, for an SBIG ST7 or ST8
camera for a 2-meter equivalent telescope with a spherical primary. This would
be adequate for time series photometry and fiber fed spectroscopy, but not wide
field imaging, so would be considered a somewhat specialized telescope. Dave
is also considering a Pressman-Carmichael Cassegrain design with an oblate
convex secondary for spherical correction and refractive elements for coma
correction. This would provide a much wider, reasonably well corrected field of
view although the secondary would not be low cost.
Mike recommended a pure truss structure, suggesting that even for smaller
telescopes such as this 2-meter, a truss structure would be superior to a
monocoque plate structure. Advantage could be taken, Mike suggested, of the
space between the four mirrors, assuming they were not jammed right on top of
one another. The trusses could extend beyond the tops of the mirrors to support
the overall mirror system.
Instead of a conventional optical telescope top end, Mike suggested a
quadrapod, with the four trusses outside the converging light, and an additional
fifth truss extending from a spider near the top of the telescope down through the
space between the mirrors. As the top end of the OTA would be very light
compared to the back end, the altitude axis would be near the top surface of the
primary mirrors.
2
Mike thought that conventional axels and bearings in altitude might be preferable
to large diameter elevation disks on both sides. He suggested, however, that we
might consider using two direct drive motors in altitude, one on each side to
avoid twisting the bottom section of the OTA.
In azimuth, we first considered using off-the-shelf slim line bearings, but if one
were to avoid fork flexures, it would be good to bring the forces straight down
from the altitude bearings to the azimuth rotation surface and hence to a
cylindrical pier. This would require about an 8-foot diameter bearing, probably
putting it out of the small telescope price range.
Instead, Mike suggested four wheels or sets of wheels rolling around on a flat
plate (track) on the top of a cylindrical pier. The four wheel sets would be placed
at the four corners of a square truss structure. Diagonals from the corners would
meet in the middle at a centering bearing.
Although a line-contact roller would provide the most surface contact area, it
would probably need to be slightly conical in shape and the track slightly tapered,
both somewhat difficult to implement. We discussed, alternatively, having one or
more “U” shaped wheels at each corner (if more than one they would need to be
on independent bearings).
We made some rough weight estimates. If the 1-meter mirrors were somewhat
thin meniscus mirrors, say 25 mm thick, they would weigh about 120 lbs. each,
and the four together would weight about 500 lbs. Mike thought that if the truss
structure was made primarily from carbon fiber tubes, that the entire telescope,
not including the mirrors, should weigh less than 500 lbs. This would give a total
weight for a 2- meter equivalent telescope of less than 1000 lbs. It could
probably be assembled by two persons without any cranes or hoists. If 55 mm
thick mirrors were used instead of 25 mm thick mirrors, the overall weight might
double to 2000 lbs. The telescope would, roughly, be 8 feet across and 24 feet
high.
I met with Chris Carter (Gemini Controls Engineer) on the following day, and over
lunch we discussed active support of primary mirrors and, later on, included Mike
Sheehan and Daigo Tomono (Subaru Engineering) in the discussion. The 182
actuators on the Subaru actually hold almost the entire weight of the Subaru’s
primary mirror (except where it very lightly rests on three defining hard points).
Thus the Subaru is magnetically levitated. Each actuator provides considerable
force and has an individual feedback loop. Gemini’s primary mirror, on the other
hand, is supported primarily by air pressure, with a seal around the periphery of
the mirror and the ability to change the air pressure versus altitude. Some 120
voice coils and permanent magnets supply the “tweaking” forces required to
bring the mirror into its desired shape.
3
Chirs Carter and Russ Genet in the Gemini electronics lab. Between them is the voice coil
mechanism that adjusts the shape of the Gemini’s 8-meter mirror. Some 120 similar coils are on
the telescope. This one is a spare.
These voice coil / magnet assemblies were made by BEI Kimco. Chris’ lab test
setup can be seen in the figure above. At 5 volts and about 1 amp of current, I
could not longer force any significant coil movement. Chris pointed out that we
could easily make our own coils. Dave Rowe has located some low cost
speakers that should work well, complete with voice coil, magnet, and centering
spider. We would just toss the speaker cone and frame.
Close up view of the voice BEI Kimco voice coil and magnet. The magnet is inside of the metal
outer sleeve, and cannot be seen. At 5 volts and 1 amp, I could no longer move the coil in and
out. Very much smaller (and lower cost) coils and magnets could be used for mirrors weighing
pounds instead of tons.
4
Chris recommended, and I think we all concurred, that it would, for smaller
telescopes, be worthwhile to consider floating the mirror on conventional
whiffletree supports and just tweaking the mirrors with the voice coil actuators at
locations between the whiffletree points. The goal here would be to just correct
the lower order Zernike terms, so this scheme would work best on mirrors of
medium flexibility, and would only require a dozen or so actuators.
Chris has considered how we might implement the control system for a dozen or
so voice coil actuators. He suggested we consider using the SPI buss, which
can be easily driven from the parallel port of a PC. Maxim Max 5352 12-bit D/A
converters are about $12 each from DigiKey. The audio/operational amplifiers
required are about the same price. If one wound one’s own coils, the cost in
electronics, magnets, and wires would be around $50 per channel. Perhaps a
unit with a dozen or two channels, power supply, etc., could be built for around
$2000.
All four of us discussed, at some length, whether or not implementation of an
active primary mirror would require some sort of on-line wavefront sensor, such
as a Shack-Hartmann sensor. Although Gemini does utilize an on-line wavefront
sensor for primary mirror adjustments, Subaru does not. Subaru calibrates their
system offline with a wavefront sensor on bright stars with 30 second integrations
to average out atmospheric effects. (Bright to them is 14th magnitude—brighter
than that they overflow their camera in 30 seconds.) However, they only need to
do this at infrequent intervals, every six to eighteen months. When they make a
calibration run, they find the sensor positions that give the best mirror figure for
every 15 degrees in altitude. During normal operation, a lookup table and
interpolation provides the settings for the actuators.
We discussed whether or not for a telescope with just a dozen or two actuators
fine tuning a mirror supported by a whiffletree, one would even need a wavefront
sensor for calibration runs, let alone regular operation. Our conclusion was that
a wavefront sensor would probably not be required. Given some time for
experimentation, a computer algorithm could be developed to automatically
search for the best settings at each of several altitudes, observing a bright star at
each altitude and making adjustments to minimize the image size. These
settings could then be stored in a lookup table and interpolated for any specific
altitude. This is not to suggest that one might not want to employ a wavefront
sensor, only to suggest that it might not be necessary.
My thanks to Mike, Chris, and Daigo for taking the time to consider how the
technology used on large alt-az telescopes might be applied to smaller aperture
telescopes.
5